3 ms·

View all pages: http://www.wired.com/wiredenterprise/2012/11/google-spanner-time/all/ http://www.wired.com/wiredenterprise/2012/11/google-spanner-... Also: > “

by Bakkot 14y ago

View all pages: http://www.wired.com/wiredenterprise/2012/11/google-spanner-time/all/ http://www.wired.com/wiredenterprise/2012/11/google-spanner-... Also: > “We can commit data at two different locations — say the West Coast [of the United States] and Europe — and still have some agreed upon ordering between them,” Fikes says, “So, if the West Coast write happens first and then the one in Europe happens, the whole system knows that — and there’s no possibility of then being viewed in a different order.” That's a large enough scale that you have to deal with relativity (light takes almost precisely 0.03 seconds to go from Palo Alto to Paris, eg). So in some sense there is no correct ordering. Anyone know how they deal with this? Have they just chosen some arbitrary point to make their reference frame, for purposes of ordering commits?

andrewcooke 14y ago
you're confusing finite speed of light and relativity - they're not really the same (although they are of course related). the relativistic sense of "no correct ordering" is irrelevant to the problem that google are solving (two observers, watching google's machines, from different, high-speed spaceships, may still disagree about the order of what happens in paris and the usa, or even in two machines in the same datacentre, but google doesn't care). they've chosen gps time as a universal clock.
whitewhim 14y ago
The entire underpinnings of GPS relies on the speed of light to calculate positions as the GPS satellites send their position, velocity and a current time stamp. These can be used to generate a distance from the satellite based on the velocity of the light. Through multiple distance readings of multiple satellites a position can be determined. Relativity does in fact play a role since the satellites are not geosynchronous they are moving in comparison to Earths reference frame. This creates non negligible errors in GPS which have been accounted for in the GPS equations. If they hadn't been we would have much more error in GPS
- andrewcooke 14y ago
  there are at least two relativistic effects involved in gps timings - http://www.aapt.org/doorway/tgru/articles/ashbyarticle.pdf http://www.aapt.org/doorway/tgru/articles/ashbyarticle.pdf - but they would not be solved if the satellites were in geosynchronous orbit as the earth itself does not have a single non-accelerating reference frame.

by Bakkot 14y ago

andrewcooke 14y ago

you're confusing finite speed of light and relativity - they're not really the same (although they are of course related). the relativistic sense of "no correct ordering" is irrelevant to the problem that google are solving (two observers, watching google's machines, from different, high-speed spaceships, may still disagree about the order of what happens in paris and the usa, or even in two machines in the same datacentre, but google doesn't care). they've chosen gps time as a universal clock.

whitewhim 14y ago

The entire underpinnings of GPS relies on the speed of light to calculate positions as the GPS satellites send their position, velocity and a current time stamp. These can be used to generate a distance from the satellite based on the velocity of the light. Through multiple distance readings of multiple satellites a position can be determined. Relativity does in fact play a role since the satellites are not geosynchronous they are moving in comparison to Earths reference frame. This creates non negligible errors in GPS which have been accounted for in the GPS equations. If they hadn't been we would have much more error in GPS

andrewcooke 14y ago
there are at least two relativistic effects involved in gps timings - http://www.aapt.org/doorway/tgru/articles/ashbyarticle.pdf http://www.aapt.org/doorway/tgru/articles/ashbyarticle.pdf - but they would not be solved if the satellites were in geosynchronous orbit as the earth itself does not have a single non-accelerating reference frame.